feat: Integrate mem0 for conversation memory - #1
Conversation
📝 WalkthroughWalkthroughThis pull request adds a standalone mem0 memory experiment script backed by Ollama, demonstrating how to initialize a Memory client with local SQLite persistence and perform sequential add/search operations on user-scoped memories. Database files are excluded from git tracking. ChangesMem0 Memory Experiment
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
ESLint skipped: no ESLint configuration detected in root package.json. To enable, add Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
🧹 Nitpick comments (1)
mem0.ts (1)
22-23: ⚡ Quick winRemove
@ts-ignoreand address the underlying type issue.The
@ts-ignoredirective suppresses type errors without resolving the root cause. Investigate whether theMemoryconstructor accepts the config structure, add proper type assertions, or install missing type definitions to improve type safety.♻️ Possible approaches to remove `@ts-ignore`
If the config type is incompatible, consider:
-// `@ts-ignore` -const memory = new Memory(config); +const memory = new Memory(config as any); // Temporary - investigate proper typesOr better, check if
mem0aiprovides type definitions:-// `@ts-ignore` -const memory = new Memory(config); +// Install `@types/mem0ai` if available, or define proper interface +const memory = new Memory(config);🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@mem0.ts` around lines 22 - 23, Remove the "// `@ts-ignore`" by fixing the type mismatch between the `config` object and the `Memory` constructor: inspect the `Memory` class constructor signature and either reshape `config` to match that type, cast `config` with a proper interface or type from the library (e.g., import the `MemoryConfig`/`MemoryOptions` type if provided by mem0ai), or update the code to call `new Memory(...)` with the expected parameters (or a typed generic) so the TypeScript compiler is satisfied; if the library lacks typings, add or install proper type definitions (or declare a minimal interface) and use that type in the `config` declaration before constructing `Memory`.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@index.ts`:
- Around line 14-36: Centralize the Ollama host into a single config value
instead of hard-coding the LAN IP in three places: introduce a single constant
(e.g., OLLAMA_BASE_URL or read from env) and use it when calling createOllama
(replace the baseURL value used by createOllama), when creating gemma3_270m (the
ollama(...) call), and inside mem0Config at llm.config.ollama_base_url and
embedder.config.ollama_base_url so the endpoint is defined once and can be
changed per-environment.
- Around line 238-270: Wrap the memory I/O so failures are best-effort: call
memory.search inside its own try/catch (protecting the code that builds context
and the variable context/searchResults) and on error log the failure and leave
context as the default "No relevant context found."; similarly, invoke
memory.add in a fire-and-forget or try/catch pattern (e.g.,
memory.add(...).catch(...) or await inside try/catch) so any error is logged but
does not rethrow into the outer catch and thus cannot block generating or
playing the assistant response; ensure you reference the existing memory.search
and memory.add calls and the context variable when applying these changes.
- Around line 240-249: The memory.search call is using the wrong OSS contract
keys and paths; change the filter key to filters: { userId:
"voice-assistant-user" } and adapt the result handling to the Mem0 shape by
reading the results array (either searchResults.results if the call returns an
object or searchResults if it returns an array) and extracting text from each
item’s memory field (e.g., item.memory?.text) when building the context string
used in the context variable and console.log; update references to
searchResults.length to check the correct array (results) and map over results
to join memory text entries.
In `@mem0.ts`:
- Line 63: The call to the async main() at the bottom risks unhandled promise
rejections; wrap the invocation to handle rejections from main() by appending a
.catch handler (e.g., main().catch(...)) that logs the error via the existing
logger or console.error and exits with a non-zero code; update the invocation of
main() to use main().catch(err => { processLogger?.error?.('Unhandled error in
main', err) || console.error(err); process.exit(1); }) so any rejection is
surfaced and the process terminates cleanly.
In `@package.json`:
- Line 13: Replace the wildcard dependency entry "mem0ai": "*" in package.json
with the exact version this PR was validated against (use the version found in
the lockfile or the tested tag), e.g. "mem0ai": "x.y.z"; update the lockfile
(npm install or yarn install) to persist the change and run the test suite to
confirm compatibility; ensure any references relying on mem0ai behavior remain
valid for that pinned version.
---
Nitpick comments:
In `@mem0.ts`:
- Around line 22-23: Remove the "// `@ts-ignore`" by fixing the type mismatch
between the `config` object and the `Memory` constructor: inspect the `Memory`
class constructor signature and either reshape `config` to match that type, cast
`config` with a proper interface or type from the library (e.g., import the
`MemoryConfig`/`MemoryOptions` type if provided by mem0ai), or update the code
to call `new Memory(...)` with the expected parameters (or a typed generic) so
the TypeScript compiler is satisfied; if the library lacks typings, add or
install proper type definitions (or declare a minimal interface) and use that
type in the `config` declaration before constructing `Memory`.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 6df75fc8-0292-4a77-a124-55fce5324cbd
📒 Files selected for processing (3)
index.tsmem0.tspackage.json
| const ollama = createOllama({ | ||
| baseURL: "http://192.168.1.4:11434/api", | ||
| baseURL: "http://192.168.1.6:11434/api", | ||
| }); | ||
| const gemma3_270m = ollama("gemma3:270m"); | ||
|
|
||
| // Initialize Mem0 | ||
| const mem0Config = { | ||
| llm: { | ||
| provider: "ollama", | ||
| config: { | ||
| model: "gemma3:270m", | ||
| ollama_base_url: "http://192.168.1.6:11434/api", | ||
| }, | ||
| }, | ||
| embedder: { | ||
| provider: "ollama", | ||
| config: { | ||
| model: "embeddinggemma:300m", | ||
| ollama_base_url: "http://192.168.1.6:11434/api", | ||
| embeddingDims: 128, // 768, with smaller options available (512, 256, or 128) | ||
| }, | ||
| }, | ||
| }; |
There was a problem hiding this comment.
Move the Ollama host into config instead of baking in one LAN IP.
Lines 15, 25, and 32 tie both the assistant and Mem0 to a single private address. That makes this feature fail anywhere outside that network and leaves the endpoint duplicated in three places.
🔧 Proposed fix
+const OLLAMA_BASE_URL =
+ process.env.OLLAMA_BASE_URL ?? "http://127.0.0.1:11434/api";
+
const ollama = createOllama({
- baseURL: "http://192.168.1.6:11434/api",
+ baseURL: OLLAMA_BASE_URL,
});
@@
model: "gemma3:270m",
- ollama_base_url: "http://192.168.1.6:11434/api",
+ ollama_base_url: OLLAMA_BASE_URL,
@@
model: "embeddinggemma:300m",
- ollama_base_url: "http://192.168.1.6:11434/api",
+ ollama_base_url: OLLAMA_BASE_URL,📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| const ollama = createOllama({ | |
| baseURL: "http://192.168.1.4:11434/api", | |
| baseURL: "http://192.168.1.6:11434/api", | |
| }); | |
| const gemma3_270m = ollama("gemma3:270m"); | |
| // Initialize Mem0 | |
| const mem0Config = { | |
| llm: { | |
| provider: "ollama", | |
| config: { | |
| model: "gemma3:270m", | |
| ollama_base_url: "http://192.168.1.6:11434/api", | |
| }, | |
| }, | |
| embedder: { | |
| provider: "ollama", | |
| config: { | |
| model: "embeddinggemma:300m", | |
| ollama_base_url: "http://192.168.1.6:11434/api", | |
| embeddingDims: 128, // 768, with smaller options available (512, 256, or 128) | |
| }, | |
| }, | |
| }; | |
| const OLLAMA_BASE_URL = | |
| process.env.OLLAMA_BASE_URL ?? "http://127.0.0.1:11434/api"; | |
| const ollama = createOllama({ | |
| baseURL: OLLAMA_BASE_URL, | |
| }); | |
| const gemma3_270m = ollama("gemma3:270m"); | |
| // Initialize Mem0 | |
| const mem0Config = { | |
| llm: { | |
| provider: "ollama", | |
| config: { | |
| model: "gemma3:270m", | |
| ollama_base_url: OLLAMA_BASE_URL, | |
| }, | |
| }, | |
| embedder: { | |
| provider: "ollama", | |
| config: { | |
| model: "embeddinggemma:300m", | |
| ollama_base_url: OLLAMA_BASE_URL, | |
| embeddingDims: 128, // 768, with smaller options available (512, 256, or 128) | |
| }, | |
| }, | |
| }; |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@index.ts` around lines 14 - 36, Centralize the Ollama host into a single
config value instead of hard-coding the LAN IP in three places: introduce a
single constant (e.g., OLLAMA_BASE_URL or read from env) and use it when calling
createOllama (replace the baseURL value used by createOllama), when creating
gemma3_270m (the ollama(...) call), and inside mem0Config at
llm.config.ollama_base_url and embedder.config.ollama_base_url so the endpoint
is defined once and can be changed per-environment.
| // 3. Search for relevant memories | ||
| console.log("🧠 Searching memories..."); | ||
| const searchResults = await memory.search(userText, { | ||
| filters: { user_id: "voice-assistant-user" }, | ||
| }); | ||
| let context = "No relevant context found."; | ||
| if (searchResults && searchResults.length > 0) { | ||
| context = | ||
| "Here is some relevant context from our previous conversations:\n" + | ||
| searchResults.map((mem: any) => mem.text).join("\n"); | ||
| } | ||
| console.log(`📝 Context: "${context}"`); | ||
|
|
||
| // 4. Get LLM response | ||
| console.log("🤖 Generating response..."); | ||
| const prompt = `You are a helpful voice assistant. Keep responses concise and conversational and very short, one or two sentences. | ||
|
|
||
| ${context} | ||
|
|
||
| Based on the context above, answer the user's new message: "${userText}"`; | ||
|
|
||
| const { text: llmResponse } = await generateText({ | ||
| model: gemma3_270m, | ||
| prompt: `You are a helpful voice assistant. Keep responses concise and conversational and very short, one or two sentences. User said: "${userText}"` | ||
| prompt: prompt, | ||
| }); | ||
|
|
||
| console.log(`🤖 Assistant: "${llmResponse}"`); | ||
|
|
||
| // 4. Convert response to speech and play | ||
|
|
||
| // 5. Add conversation to memory | ||
| console.log("💾 Adding conversation to memory..."); | ||
| await memory.add(`User: ${userText}\nAssistant: ${llmResponse}`, { | ||
| userId: "voice-assistant-user", | ||
| }); |
There was a problem hiding this comment.
Make memory I/O best-effort so it can't block replies.
Right now any memory.search() or memory.add() failure drops into the outer catch. A transient Mem0/Ollama issue means no answer gets generated or no audio gets played, even though the assistant could continue without memory.
🛡️ Proposed fix
- console.log("🧠 Searching memories...");
- const searchResponse = await memory.search(userText, {
- filters: { userId: "voice-assistant-user" },
- });
let context = "No relevant context found.";
- const searchResults = searchResponse.results ?? [];
- if (searchResults.length > 0) {
- context =
- "Here is some relevant context from our previous conversations:\n" +
- searchResults.map((mem: any) => mem.memory).join("\n");
- }
+ try {
+ console.log("🧠 Searching memories...");
+ const searchResponse = await memory.search(userText, {
+ filters: { userId: "voice-assistant-user" },
+ });
+ const searchResults = searchResponse.results ?? [];
+ if (searchResults.length > 0) {
+ context =
+ "Here is some relevant context from our previous conversations:\n" +
+ searchResults.map((mem: any) => mem.memory).join("\n");
+ }
+ } catch (error) {
+ console.warn("⚠️ Memory search failed, continuing without context:", error);
+ }
@@
- console.log("💾 Adding conversation to memory...");
- await memory.add(`User: ${userText}\nAssistant: ${llmResponse}`, {
- userId: "voice-assistant-user",
- });
+ try {
+ console.log("💾 Adding conversation to memory...");
+ await memory.add(`User: ${userText}\nAssistant: ${llmResponse}`, {
+ userId: "voice-assistant-user",
+ });
+ } catch (error) {
+ console.warn("⚠️ Memory write failed, continuing without persistence:", error);
+ }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@index.ts` around lines 238 - 270, Wrap the memory I/O so failures are
best-effort: call memory.search inside its own try/catch (protecting the code
that builds context and the variable context/searchResults) and on error log the
failure and leave context as the default "No relevant context found.";
similarly, invoke memory.add in a fire-and-forget or try/catch pattern (e.g.,
memory.add(...).catch(...) or await inside try/catch) so any error is logged but
does not rethrow into the outer catch and thus cannot block generating or
playing the assistant response; ensure you reference the existing memory.search
and memory.add calls and the context variable when applying these changes.
| const searchResults = await memory.search(userText, { | ||
| filters: { user_id: "voice-assistant-user" }, | ||
| }); | ||
| let context = "No relevant context found."; | ||
| if (searchResults && searchResults.length > 0) { | ||
| context = | ||
| "Here is some relevant context from our previous conversations:\n" + | ||
| searchResults.map((mem: any) => mem.text).join("\n"); | ||
| } | ||
| console.log(`📝 Context: "${context}"`); |
There was a problem hiding this comment.
Use the TypeScript OSS Mem0 search contract here.
The TypeScript OSS examples use filters: { userId: ... }, and the Node quickstart shows search() returning a results array whose items expose memory, not text. With the current user_id / length / mem.text path, this branch never builds real retrieval context for the prompt. (docs.mem0.ai)
🧠 Proposed fix
- const searchResults = await memory.search(userText, {
- filters: { user_id: "voice-assistant-user" },
- });
+ const searchResponse = await memory.search(userText, {
+ filters: { userId: "voice-assistant-user" },
+ });
let context = "No relevant context found.";
- if (searchResults && searchResults.length > 0) {
+ const searchResults = searchResponse.results ?? [];
+ if (searchResults.length > 0) {
context =
"Here is some relevant context from our previous conversations:\n" +
- searchResults.map((mem: any) => mem.text).join("\n");
+ searchResults.map((mem: any) => mem.memory).join("\n");
}📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| const searchResults = await memory.search(userText, { | |
| filters: { user_id: "voice-assistant-user" }, | |
| }); | |
| let context = "No relevant context found."; | |
| if (searchResults && searchResults.length > 0) { | |
| context = | |
| "Here is some relevant context from our previous conversations:\n" + | |
| searchResults.map((mem: any) => mem.text).join("\n"); | |
| } | |
| console.log(`📝 Context: "${context}"`); | |
| const searchResponse = await memory.search(userText, { | |
| filters: { userId: "voice-assistant-user" }, | |
| }); | |
| let context = "No relevant context found."; | |
| const searchResults = searchResponse.results ?? []; | |
| if (searchResults.length > 0) { | |
| context = | |
| "Here is some relevant context from our previous conversations:\n" + | |
| searchResults.map((mem: any) => mem.memory).join("\n"); | |
| } | |
| console.log(`📝 Context: "${context}"`); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@index.ts` around lines 240 - 249, The memory.search call is using the wrong
OSS contract keys and paths; change the filter key to filters: { userId:
"voice-assistant-user" } and adapt the result handling to the Mem0 shape by
reading the results array (either searchResults.results if the call returns an
object or searchResults if it returns an array) and extracting text from each
item’s memory field (e.g., item.memory?.text) when building the context string
used in the context variable and console.log; update references to
searchResults.length to check the correct array (results) and map over results
to join memory text entries.
| } | ||
| } | ||
|
|
||
| main(); |
There was a problem hiding this comment.
Handle promise rejection from main() invocation.
The async main() function is invoked without handling potential promise rejections. While the function has internal try/catch, uncaught errors during promise creation could result in unhandled rejections.
🛡️ Proposed fix to handle promise rejection
-main();
+main().catch((error) => {
+ console.error("Fatal error:", error);
+ process.exit(1);
+});📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| main(); | |
| main().catch((error) => { | |
| console.error("Fatal error:", error); | |
| process.exit(1); | |
| }); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@mem0.ts` at line 63, The call to the async main() at the bottom risks
unhandled promise rejections; wrap the invocation to handle rejections from
main() by appending a .catch handler (e.g., main().catch(...)) that logs the
error via the existing logger or console.error and exits with a non-zero code;
update the invocation of main() to use main().catch(err => {
processLogger?.error?.('Unhandled error in main', err) || console.error(err);
process.exit(1); }) so any rejection is surfaced and the process terminates
cleanly.
| "description": "", | ||
| "dependencies": { | ||
| "@huggingface/transformers": "^3.7.2", | ||
| "mem0ai": "*", |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# Find and read package.json
git ls-files package.jsonRepository: U1F30C/oasis
Length of output: 69
🏁 Script executed:
# Read the package.json file to verify the mem0ai dependency and compare with others
cat -n package.json | head -40Repository: U1F30C/oasis
Length of output: 808
🏁 Script executed:
# Check for lockfiles that might mitigate non-reproducibility concerns
ls -la | grep -E 'package-lock|yarn\.lock|pnpm-lock'Repository: U1F30C/oasis
Length of output: 124
🌐 Web query:
mem0ai sdk changelog recent API changes breaking changes
💡 Result:
Mem0 has undergone significant API changes recently, particularly with the release of v2.0.0 for the Python SDK and v3.0.0 for the Node SDK [1][2]. These updates introduced several breaking changes aimed at standardizing the API and improving performance [3][2]. Key Breaking Changes: 1. Filter-Based Search: Top-level entity parameters (user_id, agent_id, run_id) are no longer supported in methods like search [3][4]. These must now be passed within a filters dictionary [3][4]. For example, instead of client.search(query, user_id="alice"), you must use client.search(query, filters={"user_id": "alice"}) [4]. 2. Parameter Casing: In the Node SDK, all request parameters and response fields have been standardized to camelCase (e.g., user_id becomes userId, top_k becomes topK) [2]. 3. Graph Memory Removal: Graph memory support has been removed from the OSS SDK [3][2]. Users should remove enable_graph and graph_store configurations [3]. Graph features are now handled at the project level via the Platform API [3]. 4. Default Parameter Changes: - The default top_k for search and get_all has been reduced from 100 to 20 [3][2]. - The default threshold for search is now 0.1 [3]. - The default rerank behavior in search is now False [3]. 5. Constructor Cleanup: organizationId, projectId, and related fields have been removed from the MemoryClient constructor [3][2]. 6. Validation: Stricter validation has been introduced; for instance, empty-string or whitespace-only entity IDs now raise a ValueError [3]. For detailed migration instructions, refer to the official Mem0 documentation [3][5].
Citations:
- 1: https://github.com/mem0ai/mem0/releases/tag/v2.0.2
- 2: https://github.com/mem0ai/mem0/releases/tag/ts-v3.0.0
- 3: https://docs.mem0.ai/changelog/sdk
- 4: [Bug] Mem0 2.0 API breaking change causes Mem0LongTermMemory to fail in AgentScope agentscope-ai/agentscope#1511
- 5: https://docs.mem0.ai/migration/api-changes
Pin mem0ai to the version this PR was validated against.
Line 13 uses "*", which is inconsistent with all other dependencies. While the lockfile ensures reproducible installs, using a wildcard with mem0ai creates a maintenance risk: the Node SDK released v3.0.0 with breaking changes including parameter casing (user_id → userId), filter structure, and modified defaults. Pinning to a known working version prevents unexpected breaks during dependency updates.
📦 Proposed fix
- "mem0ai": "*",
+ "mem0ai": "^<validated-version>",🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@package.json` at line 13, Replace the wildcard dependency entry "mem0ai": "*"
in package.json with the exact version this PR was validated against (use the
version found in the lockfile or the tested tag), e.g. "mem0ai": "x.y.z"; update
the lockfile (npm install or yarn install) to persist the change and run the
test suite to confirm compatibility; ensure any references relying on mem0ai
behavior remain valid for that pinned version.
03a8ddd to
1440b9a
Compare
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
experiments/mem0.ts (1)
55-56: ⚡ Quick winReplace
@ts-ignorewithsatisfies Partial<MemoryConfig>for better type safety.The
@ts-ignoreon line 55 suppresses type checking at the dependency boundary. SinceMemoryConfigis exported from mem0ai/oss, you can import it and apply thesatisfiesoperator to the config object instead:Suggested change
-import { Memory } from "mem0ai/oss"; +import { Memory } from "mem0ai/oss"; +import type { MemoryConfig } from "mem0ai/oss"; const config = { llm: { @@ -50,8 +51,7 @@ const config = { }, }, -}; +} satisfies Partial<MemoryConfig>; -// `@ts-ignore` — config is Partial<MemoryConfig>, TS doesn't narrow it perfectly const memory = new Memory(config);This maintains full type safety on config keys while eliminating the type suppression.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@experiments/mem0.ts` around lines 55 - 56, Replace the `@ts-ignore` by importing the MemoryConfig type and using the TypeScript satisfies operator on the config object: import MemoryConfig from the mem0ai/oss package (or add a named import if exported) and declare the config variable as "const config = { ... } satisfies Partial<MemoryConfig>" so you can remove the `@ts-ignore` and then call new Memory(config); update the top imports to include MemoryConfig and ensure the config object shape still conforms to Partial<MemoryConfig>.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@experiments/mem0.ts`:
- Around line 91-93: The catch block currently only logs the error
(console.error("Error:", err)); update it to set a non-zero exit code so
failures propagate to CI: inside the catch handle the same err variable but also
set process.exitCode = 1 (or call process.exit(1)) after logging; ensure this
change is made in the top-level try/catch where console.error("Error:", err)
appears so the script returns a failing status when the experiment fails.
- Line 10: Guard against a missing OLLAMA_BASE_URL before calling .replace():
remove the module-level non-null assertion and either compute OLLAMA_HOST inside
main() after validating process.env.OLLAMA_BASE_URL or use a safe expression
(e.g., check process.env.OLLAMA_BASE_URL exists and then call .replace or
provide a default) so that the code that defines OLLAMA_HOST (symbol:
OLLAMA_HOST) does not call .replace() on undefined; ensure any error is handled
or thrown with a clear message referencing OLLAMA_BASE_URL.
---
Nitpick comments:
In `@experiments/mem0.ts`:
- Around line 55-56: Replace the `@ts-ignore` by importing the MemoryConfig type
and using the TypeScript satisfies operator on the config object: import
MemoryConfig from the mem0ai/oss package (or add a named import if exported) and
declare the config variable as "const config = { ... } satisfies
Partial<MemoryConfig>" so you can remove the `@ts-ignore` and then call new
Memory(config); update the top imports to include MemoryConfig and ensure the
config object shape still conforms to Partial<MemoryConfig>.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: ce2e0281-e4ca-41eb-ae92-35dd0156dde2
📒 Files selected for processing (2)
.gitignoreexperiments/mem0.ts
✅ Files skipped from review due to trivial changes (1)
- .gitignore
| // 4. search() options use `filters: { user_id }`, not top-level `userId` | ||
|
|
||
| // mem0's ollama client appends /api itself — strip it if OLLAMA_BASE_URL has it | ||
| const OLLAMA_HOST = process.env.OLLAMA_BASE_URL!.replace(/\/api$/, ""); |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
find . -name "mem0.ts" -type fRepository: U1F30C/oasis
Length of output: 78
🏁 Script executed:
cat -n ./experiments/mem0.tsRepository: U1F30C/oasis
Length of output: 3563
🏁 Script executed:
rg "OLLAMA_BASE_URL" ./experiments/mem0.tsRepository: U1F30C/oasis
Length of output: 209
🏁 Script executed:
rg -i "process.env" ./experiments/mem0.tsRepository: U1F30C/oasis
Length of output: 128
🏁 Script executed:
rg "OLLAMA" . --type ts --type js | head -20Repository: U1F30C/oasis
Length of output: 413
🏁 Script executed:
cat -n ./src/server/llm.ts | head -50Repository: U1F30C/oasis
Length of output: 478
🏁 Script executed:
find . -name ".env*" -o -name "*.env.example" | head -10Repository: U1F30C/oasis
Length of output: 71
🏁 Script executed:
cat ./.env.exampleRepository: U1F30C/oasis
Length of output: 793
Guard OLLAMA_BASE_URL before calling .replace().
Line 10 executes at module load time before the try/catch in main() is reached. If OLLAMA_BASE_URL is unset, the non-null assertion ! suppresses the TypeScript warning but does not prevent the runtime error; .replace() will throw a TypeError.
Suggested fix
-const OLLAMA_HOST = process.env.OLLAMA_BASE_URL!.replace(/\/api$/, "");
+const rawOllamaBaseUrl = process.env.OLLAMA_BASE_URL;
+if (!rawOllamaBaseUrl) {
+ throw new Error("Missing required env var: OLLAMA_BASE_URL");
+}
+const OLLAMA_HOST = rawOllamaBaseUrl.replace(/\/api\/?$/, "");📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| const OLLAMA_HOST = process.env.OLLAMA_BASE_URL!.replace(/\/api$/, ""); | |
| const rawOllamaBaseUrl = process.env.OLLAMA_BASE_URL; | |
| if (!rawOllamaBaseUrl) { | |
| throw new Error("Missing required env var: OLLAMA_BASE_URL"); | |
| } | |
| const OLLAMA_HOST = rawOllamaBaseUrl.replace(/\/api\/?$/, ""); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@experiments/mem0.ts` at line 10, Guard against a missing OLLAMA_BASE_URL
before calling .replace(): remove the module-level non-null assertion and either
compute OLLAMA_HOST inside main() after validating process.env.OLLAMA_BASE_URL
or use a safe expression (e.g., check process.env.OLLAMA_BASE_URL exists and
then call .replace or provide a default) so that the code that defines
OLLAMA_HOST (symbol: OLLAMA_HOST) does not call .replace() on undefined; ensure
any error is handled or thrown with a clear message referencing OLLAMA_BASE_URL.
| } catch (err) { | ||
| console.error("Error:", err); | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's explore the repository structure
git ls-files | head -20Repository: U1F30C/oasis
Length of output: 973
🏁 Script executed:
# Look for experiments/mem0.ts
fd mem0.tsRepository: U1F30C/oasis
Length of output: 76
🏁 Script executed:
# Check if the file exists and its size
wc -l experiments/mem0.ts 2>/dev/null || echo "File not found, searching..."
fd -type f -name "*.ts" | grep -i memRepository: U1F30C/oasis
Length of output: 268
🏁 Script executed:
# Read the file to see the catch block context
cat -n experiments/mem0.ts | sed -n '80,96p'Repository: U1F30C/oasis
Length of output: 666
🏁 Script executed:
# Also check the beginning to understand the file structure
head -30 experiments/mem0.ts | cat -nRepository: U1F30C/oasis
Length of output: 1116
Set a non-zero exit code when the experiment fails.
The catch block at lines 91-93 only logs errors without setting an exit code. Automation systems and CI/CD pipelines will treat the script as successful (exit code 0) even when it fails, preventing proper error detection.
Suggested fix
} catch (err) {
console.error("Error:", err);
+ process.exitCode = 1;
}📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| } catch (err) { | |
| console.error("Error:", err); | |
| } | |
| } catch (err) { | |
| console.error("Error:", err); | |
| process.exitCode = 1; | |
| } |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@experiments/mem0.ts` around lines 91 - 93, The catch block currently only
logs the error (console.error("Error:", err)); update it to set a non-zero exit
code so failures propagate to CI: inside the catch handle the same err variable
but also set process.exitCode = 1 (or call process.exit(1)) after logging;
ensure this change is made in the top-level try/catch where
console.error("Error:", err) appears so the script returns a failing status when
the experiment fails.
1 participant
This commit introduces the mem0.ai library to provide conversational memory to the voice assistant.
mem0aidependency topackage.json.mem0.tsto demonstrate and test the mem0 integration with Ollama for both LLM and embeddings, as per user specification.index.ts.Summary by CodeRabbit
Chores
Tests